Effort Estimation using Neural Network and Metaheuristic Optimizer

One of the most essential factors in software development is to have an accurate forecast of workforce. In a project, the project management team must ensure accurate effort as well cost estimation in order to develop a high quality project which satisfies the customer's requirements. Inaccuracy in effort estimation of a project may lead to rejection of its development. Prognosis of effort forecasting involves prediction goals, different mechanisms for measurements, updating and refining techniques, and the processes used for execution. In the course of past few years, researches have examined various distinct problems for effort forecasting and have given newer models and techniques to minimize the errors involved. A lot of different models and approaches were surveyed to find out the trend in effort estimation and it was concluded that meta-heuristic-based models perform better as compared to other techniques proposed. So, in order to find a technique that can improve the effort accuracy, the work presented in this paper proposes a Neural Network based model which performs weight optimization using a meta-heuristic algorithm called Dragonfly algorithm. Performance of the proposed model had been evaluated using COCOMO81, Albrecht and Desharnais datasets and its performance had been compared against the effort estimation models developed using traditional machine learning algorithms viz. Random Forest and Regression Tree. The values of the performance parameters show the superior performance of the proposed meta-heuristic model as compared to the machine learning models.